Dandelion is an execution exchange for compute that lowers LLM inference costs while developers and companies keep their models of choice. Compatible endpoints include OpenAI Responses, Chat Completions and Anthropic Messages for supported models. Dandelion earns a fee only when customers save, taken from savings against its retail baseline.